Reducing damage in gallium nitride inductively coupled plasma etch

University of California Santa Barbara in the USA and Ecole Polytechnique in France have developed a low-damage dry etch for III-nitride semiconductors that was effective to within 71nm of an active region [Joseph G Nedy et al, Semicond. Sci. Technol., vol30, p085019, 2015]. The process was used on test structures and for creating an air-gap cladding in a laser diode (LD) as an alternative to complex aluminium gallium nitride/gallium nitride (AlGaN/GaN) superlattice (SL) structures.

Dry etching with plasma causes damage with ion bombardment. In many compound semiconductor processes, wet etch is preferred because it causes much less damage. However, for III-nitride semiconductors a simple wet alternative is unavailable.

The test structure consisted of two indium gallium nitride (InGaN) multiple quantum well (MQW) regions grown on double-side polished (DSP) c-plane sapphire (Figure 1). Photoluminescence (PL) from the different depth MQWs was used to analyze the damage from the dry etch procedure.

The test wafer was cleaved into 6mmx12mm pieces, and mesa stripes were defined by lithographic patterning on the silicon nitride (SiNx) top-layer hard mask. The pattern was transferred to the SiNxwith inductively coupled plasma (ICP) etch with trifluoromethane/tetrafluoromethane. The power was low so that the underlying GaN served as an etch stop.

The GaN etching was performed on a Unaxis VLR high-temperature ICP system. The chamber was cleaned and seasoned before the samples were etched. Care was taken to avoid etch byproducts from the silicon carrier wafer and edge effects. A pre-treatment consisting of boron trichloride etch was applied before both the seasoning and main GaN etch runs to remove native oxide and for better etch uniformity.

The GaN etch chemistry was chlorine/argon. Low power was used to minimize sub-surface damage. The ICP was generated at 200W and the forward power was 5W. The low pressure of 2mTorr avoided re-deposition of etch products on the sidewalls and to allow high aspect ratios to be achieved in future. The substrate was heated to 200°C. The etch process parameters were at the lower limit before etch rates became unmeasurable. With the given parameters the etch rate was 26nm/minute on c-plane material.

The use of two MQWs for PL creates an in-situ control, reducing problems from changes in measurement set up and the effects of thinning and surface roughening from the etch. The PL from the deep MQW (417nm) after 500nm etch was a factor of 150% times that of before. The PL from the shallow well (469nm) was therefore normalized according to the change in the PL from the deep well, assuming that it was too far away from the etch to be affected.